Intensive care life information monitoring method applying machine learning algorithm

Abstract

The invention provides an intensive care life information monitoring method applying a machine learning algorithm, and the method comprises the steps: collecting the vital sign data of a critical patient, carrying out the preprocessing of the vital sign data, obtaining the target vital sign data, providing a high-quality data basis for the further recognition and analysis of the vital sign data, and improving the recognition efficiency of the critical patient. According to the method, historical sign data and historical anomaly recognition data of historical critical patients are utilized, a machine learning algorithm is utilized, a patient sign anomaly recognition model is obtained through training and construction, the patient sign anomaly recognition model is obtained through the machine learning algorithm, the performance of the model is guaranteed, a model basis is provided for intensive care life information monitoring, and the patient sign anomaly recognition method is provided for intensive care life information monitoring. And inputting the target vital sign data into the patient sign abnormity identification model to obtain an abnormity identification result, carrying out comprehensive analysis on the abnormity identification result, and generating an early warning prompt for medical staff, thereby realizing accurate monitoring and early warning prompt for intensive care life information, and helping to timely discover abnormal conditions and potential health risks of critical patients.

Description

Technical Field

[0001] The present invention relates to the technical field of intensive care, and particularly to a method for monitoring vital information in intensive care by applying machine learning algorithms. Background Art

[0002] Patients in intensive care are usually critically ill, and vital sign changes may occur at any time. Therefore, the importance of real-time monitoring of vital information lies in timely detecting problems and dealing with them promptly.

[0003] In recent years, machine learning algorithms have been used to predict the deterioration of the condition, and personalized risk assessment has been realized in combination with electronic health records, significantly improving the monitoring efficiency.

[0004] How to use machine learning algorithms to achieve accurate monitoring and early warning of vital information in intensive care is of great significance for timely detecting potential health risks of critically ill patients. Summary of the Invention

[0005] The present invention provides a method for monitoring vital information in intensive care by applying machine learning algorithms to solve the problems raised in the background art.

[0006] A method for monitoring vital information in intensive care by applying machine learning algorithms includes:

[0007] S1: Collect vital sign data of critically ill patients, preprocess the vital sign data to obtain target vital sign data;

[0008] S2: Based on the historical sign data and historical abnormal identification data of historical critically ill patients, use machine learning algorithms to train and construct a patient sign abnormal identification model;

[0009] S3: Input the target vital sign data into the patient sign abnormal identification model to obtain an abnormal identification result;

[0010] S4: Comprehensively analyze the abnormal identification result to generate a warning reminder for medical staff.

[0011] Preferably, in S1, collecting vital sign data of critically ill patients includes:

[0012] Collect monitoring data of critically ill patients based on sensors;

[0013] Based on real-time transmission technology, transmit the monitoring data to the monitoring center in real time;

[0014] The monitoring center receives and stores the monitoring data to obtain vital sign data.

[0015] Preferably, in S1, preprocessing the vital sign data to obtain target vital sign data includes:

[0016] Based on the data range corresponding to the data type, eliminate error values and supplement differences for the vital sign data to obtain vital sign data with a normal data range.

[0017] Extract key information of the target features from the vital sign data with a normal data range to obtain vital sign data containing key information.

[0018] Based on the model recognition data format, perform format conversion on the vital sign data containing key information to obtain the target vital sign data.

[0019] Preferably, in S2, based on the historical vital sign data and historical abnormal recognition data of historical critically ill patients, use machine learning algorithms to train and construct a patient vital sign abnormal recognition model, including:

[0020] Mark the historical vital sign data based on the historical abnormal recognition data to obtain marked vital sign data.

[0021] Based on the marked vital sign data, train the machine learning algorithm, and construct a patient vital sign abnormal recognition model according to the training results.

[0022] Preferably, in S3, input the target vital sign data into the patient vital sign abnormal recognition model to obtain an abnormal recognition result, including:

[0023] Input the target vital sign data into the patient vital sign abnormal recognition model to obtain the abnormal data type and abnormal level.

[0024] Integrate the abnormal data type and abnormal level to obtain an abnormal recognition result.

[0025] Preferably, in S4, comprehensively analyze the abnormal recognition result to generate a warning reminder for medical staff, including:

[0026] Determine the abnormal numerical type and abnormal numerical level from the abnormal recognition result.

[0027] Perform type association based on the abnormal numerical type to obtain a type association result. Based on the type association result, combined with the abnormal numerical level, obtain an abnormal numerical weighted relationship distribution diagram.

[0028] Generate a warning reminder for medical staff based on the abnormal numerical weighted relationship distribution diagram.

[0029] Preferably, performing type association based on the abnormal numerical type to obtain a type association result. Based on the type association result, combined with the abnormal numerical level, obtain an abnormal numerical weighted relationship distribution diagram, including:

[0030] Based on historical physical sign data, obtain the first abnormal numerical type and the first occurrence frequency that occur simultaneously in the same time period, and determine the first correlation value of the first abnormal numerical type based on the first occurrence frequency of the first abnormal numerical type;

[0031] Based on historical feature description data, obtain the second abnormal numerical type and the second occurrence frequency that occur in successive time periods, and determine the second correlation value of the second abnormal numerical type based on the second occurrence frequency of the second abnormal numerical type;

[0032] Based on the first correlation value and the second correlation value, establish a type correlation rule, and perform type correlation on the abnormal numerical type based on the weighted correlation rule to obtain a type correlation result;

[0033] Based on the abnormal level of the first abnormal numerical type, perform level weighting on the first correlation value to obtain a first level weighted value, and based on the abnormal level of the second abnormal numerical type, perform level weighting on the second correlation value to determine a second level weighted value;

[0034] Based on the first level weighted value and the second level weighted value, and in combination with the abnormal numerical level, perform level weighting on the type correlation result to obtain a weighted correlation result;

[0035] Based on the weighted correlation result, obtain an initial abnormal numerical weighted relationship distribution diagram;

[0036] Based on the personalized information of critically ill patients, determine the personalized weight for the abnormal type, and perform personalized weighting on the initial abnormal numerical weighted relationship distribution diagram based on the personalized weight to obtain an abnormal numerical weighted relationship distribution diagram.

[0037] Preferably, generating a warning reminder for medical staff based on the abnormal numerical weighted relationship distribution diagram includes:

[0038] Obtain the abnormal type level, weighted correlation value, and distribution range in the abnormal numerical weighted relationship distribution diagram;

[0039] Judge whether the abnormal type level is within a preset range;

[0040] If so, based on the abnormal type, in combination with the weighted correlation value and the distribution range, determine the potential health risk, set a low warning level for the potential health risk, and generate a warning reminder for medical staff;

[0041] Otherwise, based on the abnormal type, in combination with the weighted correlation value and the distribution range, determine the health abnormal information, set a high warning level for the potential health risk, and generate a warning reminder for medical staff.

[0042] Preferably, it further includes real-time optimization of the patient physical sign abnormal recognition model, including:

[0043] Obtain the abnormal prediction results of the patient physical sign abnormality recognition model within a period of time, obtain the actual diagnosis results, and obtain the result difference between the abnormal prediction results and the actual diagnosis results;

[0044] Obtain the first physical sign data corresponding to all result differences, obtain the data characteristics of the first physical sign data, determine the proportion of the data characteristics in the historical feature data, and judge whether the proportion is greater than the preset proportion;

[0045] If so, do not adjust the historical physical sign data;

[0046] Otherwise, based on the preset proportion, adjust the historical physical sign data to obtain updated historical physical sign data;

[0047] Obtain the second physical sign data with the result difference greater than the preset result difference, obtain the historical marked features of the second physical sign data in the historical feature data, correct the historical marked features based on the result difference to obtain target marked features, and update the marks of the updated historical physical sign data based on the target marked features to obtain new training data;

[0048] Configure the first attention head of the attention mechanism for the adjusted physical sign data in the new training data, and configure the second attention head of the attention mechanism for the physical sign data of the target marked features in the new training data;

[0049] Based on the first attention head and the second attention head, add an attention mechanism inside the model, and combine the new training data to perform secondary training and optimization on the patient physical sign abnormality recognition model to obtain a real-time abnormality recognition model.

[0050] Preferably, adjusting the historical physical sign data based on the preset proportion to obtain updated historical physical sign data includes:

[0051] Adjust the proportion of the data characteristics of the first physical sign data in the historical physical sign data to be greater than the preset proportion to obtain the removed data volume;

[0052] Obtain the data content where the historical physical sign data does not appear in all result differences, and uniformly remove the data content according to the removed data volume to obtain updated historical physical sign data.

[0053] Compared with the prior art, the present invention has achieved the following beneficial effects:

[0054] By collecting the vital sign data of critically ill patients, preprocessing the vital sign data to obtain target vital sign data, providing a high-quality data basis for further identification and analysis of the vital sign data, using the historical vital sign data and historical anomaly identification data of historical critically ill patients, and using machine learning algorithms, training and constructing a patient vital sign anomaly identification model, realizing the use of machine learning algorithms to obtain a patient vital sign anomaly identification model, ensuring the performance of the model, providing a model basis for intensive care life information monitoring, inputting the target vital sign data into the patient vital sign anomaly identification model to obtain an anomaly identification result, comprehensively analyzing the anomaly identification result, generating a warning reminder for medical staff, realizing the accurate monitoring and warning reminder of intensive care life information, and helping to timely discover the abnormal conditions and potential health risks of critically ill patients.

[0055] Other features and advantages of the present invention will be described in the following specification, and part of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structure specifically pointed out in this application document.

[0056] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the accompanying drawings:

[0058] Figure 1 is a flowchart of a method for monitoring intensive care life information applying machine learning algorithms in an embodiment of the present invention;

[0059] Figure 2 is a flowchart of obtaining target vital sign data in an embodiment of the present invention;

[0060] Figure 3 is a flowchart of obtaining a patient vital sign anomaly identification model in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0062] Embodiment 1:

[0063] An embodiment of the present invention provides a method for monitoring intensive care life information applying machine learning algorithms, as Figure 1 shown, including:

[0064] S1: Collect the vital sign data of critically ill patients, preprocess the vital sign data to obtain the target vital sign data;

[0065] S2: Based on the historical vital sign data and historical abnormal identification data of historical critically ill patients, use machine learning algorithms to train and construct a patient vital sign abnormal identification model;

[0066] S3: Input the target vital sign data into the patient vital sign abnormal identification model to obtain the abnormal identification result;

[0067] S4: Conduct a comprehensive analysis of the abnormal identification result to generate a warning reminder for medical staff.

[0068] In this embodiment, the vital sign data includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, body temperature, etc.

[0069] In this embodiment, preprocessing the vital sign data includes data cleaning, data standardization, etc.

[0070] In this embodiment, the historical abnormal identification data is, for example, the determination of the abnormality of the heart rate and the corresponding abnormal level.

[0071] In this embodiment, generating a warning reminder for medical staff includes abnormal information, potential anti-risk information, and warning prompts.

[0072] The beneficial effects of the above design are as follows: By collecting the vital sign data of critically ill patients, preprocessing the vital sign data to obtain the target vital sign data, it provides a high-quality data basis for further identification and analysis of the vital sign data. Using the historical vital sign data and historical abnormal identification data of historical critically ill patients, and using machine learning algorithms to train and construct a patient vital sign abnormal identification model, it realizes obtaining the patient vital sign abnormal identification model using machine learning algorithms, ensures the performance of the model, and provides a model basis for intensive care life information monitoring. Inputting the target vital sign data into the patient vital sign abnormal identification model to obtain the abnormal identification result, and conducting a comprehensive analysis of the abnormal identification result to generate a warning reminder for medical staff, it realizes the precise monitoring and warning reminder of intensive care life information, and helps to timely detect the abnormal conditions and potential health risks of critically ill patients.

[0073] Embodiment 2:

[0074] Based on Embodiment 1, the embodiment of the present invention provides a method for monitoring intensive care life information using machine learning algorithms. In S1, collecting the vital sign data of critically ill patients includes:

[0075] Collect the monitoring data of critically ill patients based on sensors;

[0076] Based on real-time transmission technology, the monitoring data is transmitted to the monitoring center in real time;

[0077] The monitoring center receives and stores the monitoring data to obtain vital sign data.

[0078] The beneficial effects of the above design scheme are as follows: By collecting the monitoring data of critically ill patients based on sensors, and based on real-time transmission technology, the monitoring data is transmitted to the monitoring center in real time. The monitoring center receives and stores the monitoring data to obtain vital sign data, realizing real-time monitoring of the vital sign data of critically ill patients and providing time-sensitive data for intensive care life information monitoring.

[0079] Embodiment 3:

[0080] Based on Embodiment 1, an intensive care life information monitoring method applying a machine learning algorithm is provided in an embodiment of the present invention. As Figure 2 shown, in the S1, the vital sign data is preprocessed to obtain target vital sign data, including:

[0081] Based on the data range corresponding to the data type, the error values are removed and the differences are supplemented for the vital sign data to obtain vital sign data with a normal data range;

[0082] The key information of the target features is extracted from the vital sign data with a normal data range to obtain vital sign data containing the key information;

[0083] Based on the model to identify the data format, the format of the vital sign data containing the key information is converted to obtain the target vital sign data.

[0084] In this embodiment, the interference of error values is removed from the vital sign data with a normal data range.

[0085] In this embodiment, the redundant data is removed from the vital sign data containing the key information to ensure the quality of the data content.

[0086] The beneficial effects of the above design scheme are as follows: By removing the error values and supplementing the differences for the vital sign data based on the data range corresponding to the data type, vital sign data with a normal data range is obtained. The key information of the target features is extracted from the vital sign data with a normal data range to obtain vital sign data containing the key information. Based on the model to identify the data format, the format of the vital sign data containing the key information is converted to obtain the target vital sign data, providing a high-quality data basis for further identification and analysis of the vital sign data.

[0087] Embodiment 4:

[0088] Based on Embodiment 1, an embodiment of the present invention provides a method for monitoring critical care life information applying a machine learning algorithm, as follows Figure 3 As shown, in S2, based on the historical physical sign data and historical anomaly recognition data of historical critical patients, a patient physical sign anomaly recognition model is trained and constructed by using a machine learning algorithm, including:

[0089] Mark the historical physical sign data based on the historical anomaly recognition data to obtain marked physical sign data;

[0090] Based on the marked physical sign data, train the machine learning algorithm, and construct a patient physical sign anomaly recognition model according to the training results.

[0091] In this embodiment, marking the historical physical sign data means marking the abnormal data.

[0092] The beneficial effect of the above design is that by marking the historical physical sign data based on the historical anomaly recognition data to obtain marked physical sign data, training the machine learning algorithm based on the marked physical sign data, and constructing a patient physical sign anomaly recognition model according to the training results, it is realized to obtain a patient physical sign anomaly recognition model by using a machine learning algorithm, ensuring the performance of the model and providing a model basis for monitoring critical care life information.

[0093] Embodiment 5:

[0094] Based on Embodiment 1, an embodiment of the present invention provides a method for monitoring critical care life information applying a machine learning algorithm. In S3, input the target physical sign data into the patient physical sign anomaly recognition model to obtain an anomaly recognition result, including:

[0095] Input the target physical sign data into the patient physical sign anomaly recognition model to obtain the anomaly data type and anomaly level;

[0096] Integrate the anomaly data type and anomaly level to obtain an anomaly recognition result.

[0097] The beneficial effect of the above design is that by inputting the target physical sign data into the patient physical sign anomaly recognition model to obtain the anomaly data type and anomaly level, and integrating the anomaly data type and anomaly level to obtain an anomaly recognition result, it is realized to identify abnormal data by using the model.

[0098] Embodiment 6:

[0099] Based on Embodiment 1, an embodiment of the present invention provides a method for monitoring critical care life information applying a machine learning algorithm. In S4, comprehensively analyze the anomaly recognition result to generate a warning reminder for medical staff, including:

[0100] Determine the abnormal numerical type and abnormal numerical level from the abnormal recognition results;

[0101] Perform type association based on the abnormal numerical type to obtain a type association result, and based on the type association result, combine the abnormal numerical level to obtain an abnormal numerical weighted relationship distribution map;

[0102] Generate a warning reminder for medical staff based on the abnormal numerical weighted relationship distribution map.

[0103] Embodiment 7:

[0104] Based on Embodiment 6, an intensive care life information monitoring method applying a machine learning algorithm is provided in an embodiment of the present invention. Perform type association based on the abnormal numerical type to obtain a type association result, and based on the type association result, combine the abnormal numerical level to obtain an abnormal numerical weighted relationship distribution map, including:

[0105] Based on historical physical sign data, obtain the first abnormal numerical type and the first occurrence frequency that occur simultaneously in the same time period, and based on the first occurrence frequency of the first abnormal numerical type, determine the first association value of the first abnormal numerical type;

[0106] Based on historical feature description data, obtain the second abnormal numerical type and the second occurrence frequency that occur in successive time periods, and based on the second occurrence frequency of the second abnormal numerical type, determine the second association value of the second abnormal numerical type;

[0107] Based on the first association value and the second association value, establish a type association rule, and based on the weighted association rule, perform type association on the abnormal numerical type to obtain a type association result;

[0108] Based on the abnormal level of the first abnormal numerical type, perform level weighting on the first association value to obtain a first level weighted value, and based on the abnormal level of the second abnormal numerical type, perform level weighting on the second association value to determine a second level weighted value;

[0109] Based on the first level weighted value and the second level weighted value, combine the abnormal numerical level to perform level weighting on the type association result to obtain a weighted association result;

[0110] Based on the weighted association result, obtain an initial abnormal numerical weighted relationship distribution map;

[0111] Based on the personalized information of critically ill patients, determine the personalized weight for the abnormal type, and perform personalized weighting on the initial abnormal numerical weighted relationship distribution map based on the personalized weight to obtain an abnormal numerical weighted relationship distribution map.

[0112] In this embodiment, the first abnormal numerical types that occur simultaneously in the same time period are, for example, low blood oxygen and too fast heart rate occurring simultaneously, and the second abnormal numerical types that occur in successive time periods are, for example, low blood pressure and oliguria occurring within 4 hours.

[0113] In this embodiment, the higher the first occurrence frequency and the second occurrence frequency, the greater the correlation value between the two abnormal types, and key attention should be paid.

[0114] In this embodiment, the personalized information of critically ill patients is, for example, the patient's physical data and historical disease data, etc. Combining these data, weighted attention is paid to important abnormal types.

[0115] The beneficial effects of the above design scheme are as follows: By learning the correlation between abnormal types based on historical physical sign data, and performing weighted analysis on the correlation based on the abnormal level and the personalized information of critically ill patients, an initial weighted relationship distribution map of abnormal numerical values is finally obtained, and a complete, comprehensive and accurate abnormal and associated situation of critically ill patients is obtained, providing a basis for further generating warning reminders for medical staff.

[0116] Embodiment 8:

[0117] Based on Embodiment 6, an embodiment of the present invention provides a method for monitoring critical care life information using a machine learning algorithm. Generating a warning reminder for medical staff based on the weighted relationship distribution map of abnormal numerical values includes:

[0118] Obtain the abnormal type level, weighted correlation value and distribution range in the weighted relationship distribution map of abnormal numerical values;

[0119] Judge whether the abnormal type level is within a preset range;

[0120] If so, based on the abnormal type, combined with the weighted correlation value and the distribution range, determine the potential health risk, set a low warning level for the potential health risk, and generate a warning reminder for medical staff;

[0121] Otherwise, based on the abnormal type, combined with the weighted correlation value and the distribution range, determine the health abnormal information, set a high warning level for the potential health risk, and generate a warning reminder for medical staff.

[0122] In this embodiment, the weighted correlation value is the correlation value between the abnormal type and other types, and the distribution range is the distribution range of other types.

[0123] In this embodiment, when the abnormal type level of the abnormal type is high, an abnormal reminder is given, and when the abnormal type level is low, a potential risk warning is given.

[0124] The beneficial effects of the above design solution are as follows: By determining potential health risks and health abnormality information based on the weighted relationship distribution map of abnormal values, the comprehensiveness and accuracy of potential health risks and health abnormality information are ensured, and hierarchical early warnings are carried out to ensure the timeliness of early warnings, providing a basis for timely detection of abnormalities in critically ill patients.

[0125] Example 9:

[0126] Based on Example 1, the embodiment of the present invention provides a method for monitoring critical care life information applying a machine learning algorithm, further including real-time optimization of the patient's physical sign abnormality recognition model, including:

[0127] Obtain the abnormal prediction results of the patient's physical sign abnormality recognition model within a period of time, and obtain the actual diagnosis results, and obtain the result difference between the abnormal prediction results and the actual diagnosis results;

[0128] Obtain the first physical sign data corresponding to all result differences, obtain the data characteristics of the first physical sign data, determine the proportion of the data characteristics in the historical characteristic data, and judge whether the proportion is greater than the preset proportion;

[0129] If so, do not adjust the historical physical sign data;

[0130] Otherwise, based on the preset proportion, adjust the historical physical sign data to obtain updated historical physical sign data;

[0131] Obtain the second physical sign data with a result difference greater than the preset result difference, obtain the historical marked characteristics of the second physical sign data in the historical characteristic data, correct the historical marked characteristics based on the result difference to obtain target marked characteristics, and update the marking of the updated historical physical sign data based on the target marked characteristics to obtain new training data;

[0132] Configure the first attention head of the attention mechanism for the adjusted physical sign data in the new training data, and configure the second attention head of the attention mechanism for the physical sign data of the target marked characteristics in the new training data;

[0133] Based on the first attention head and the second attention head, add an attention mechanism inside the model, and combine the new training data to perform secondary training and optimization on the patient's physical sign abnormality recognition model to obtain a real-time abnormality recognition model.

[0134] The beneficial effects of the above design solution are as follows: By adjusting the data content proportion of the training data and the marking results, and adding an attention mechanism to the model, the real-time optimization of the patient's physical sign abnormality recognition model is realized, ensuring the performance of the model in actual application and improving the accuracy of predicting critical care life information.

[0135] Example 10:

[0136] Based on Embodiment 9, an embodiment of the present invention provides a method for monitoring critical care life information using a machine learning algorithm. Based on a preset ratio, historical vital sign data is adjusted to obtain updated historical vital sign data, including:

[0137] Adjust the ratio of the data features of the first vital sign data in the historical vital sign data to be greater than the preset ratio to obtain the amount of data to be removed;

[0138] Obtain the data content in which there is no difference in all results in the historical vital sign data, and uniformly remove the data content according to the amount of data to be removed to obtain updated historical vital sign data.

[0139] The beneficial effects of the above design are as follows: By adjusting the ratio of the data features of the first vital sign data in the historical vital sign data to be greater than the preset ratio to obtain the amount of data to be removed, obtaining the data content in which there is no difference in all results in the historical vital sign data, and uniformly removing the data content according to the amount of data to be removed to obtain updated historical vital sign data, the rationality of the data content of the obtained updated historical vital sign data is ensured, providing a basis for ensuring the accuracy of model training.

[0140] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of this application document and its equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for monitoring critical care life information using machine learning algorithms, characterized in that, Including: S1: Collect the vital sign data of critically ill patients, preprocess the vital sign data to obtain target vital sign data; S2: Based on the historical vital sign data and historical anomaly recognition data of historical critically ill patients, use machine learning algorithms to train and construct a patient vital sign anomaly recognition model; S3: Input the target vital sign data into the patient vital sign anomaly recognition model to obtain an anomaly recognition result; S4: Conduct a comprehensive analysis of the anomaly recognition result to generate a warning reminder for medical staff.

2. The intensive care life information monitoring method applying a machine learning algorithm according to claim 1, characterized in that In S1, collecting the vital sign data of critically ill patients includes: Collect the monitoring data of critically ill patients based on sensors; Based on real-time transmission technology, transmit the monitoring data to the monitoring center in real time; The monitoring center receives and stores the monitoring data to obtain vital sign data.

3. A method for monitoring critical care life information applying a machine learning algorithm according to claim 1, characterized in that, In S1, preprocessing the vital sign data to obtain target vital sign data includes: Based on the data range corresponding to the data type, eliminate error values and supplement differences in the vital sign data to obtain vital sign data with a normal data range; Extract the key information of the target features from the vital sign data with a normal data range to obtain vital sign data containing key information; Based on the model recognition data format, convert the format of the vital sign data containing key information to obtain target vital sign data.

4. The intensive care life information monitoring method applying a machine learning algorithm according to claim 1, characterized in that, In S2, based on the historical vital sign data and historical anomaly recognition data of historical critically ill patients, using machine learning algorithms to train and construct a patient vital sign anomaly recognition model includes: Mark the historical vital sign data based on the historical anomaly recognition data to obtain marked vital sign data; Based on the marked vital sign data, train the machine learning algorithm, and construct a patient vital sign anomaly recognition model according to the training results.

5. A method for monitoring critical care life information applying a machine learning algorithm according to claim 1, characterized in that, In S3, inputting the target vital sign data into the patient vital sign anomaly recognition model to obtain an anomaly recognition result includes: Input the target vital sign data into the patient vital sign anomaly recognition model to obtain the anomaly data type and anomaly level; Integrate the anomaly data type and anomaly level to obtain an anomaly recognition result.

6. The intensive care life information monitoring method applying a machine learning algorithm according to claim 1, characterized in that In S4, conducting a comprehensive analysis of the anomaly recognition result to generate a warning reminder for medical staff includes: Determine the anomaly value type and anomaly value level from the anomaly recognition result; Based on the anomaly value type, conduct type association to obtain a type association result. Based on the type association result, combined with the anomaly value level, obtain an anomaly value weighted relationship distribution map; Based on the anomaly value weighted relationship distribution map, generate a warning reminder for medical staff.

7. The intensive care life information monitoring method using a machine learning algorithm according to claim 6, characterized in that, Based on the anomaly value type, conduct type association to obtain a type association result. Based on the type association result, combined with the anomaly value level, obtain an anomaly value weighted relationship distribution map, including: Based on the historical vital sign data, obtain the first anomaly value type and the first occurrence frequency that occur simultaneously in the same time period. Based on the first occurrence frequency of the first anomaly value type, determine the first association value of the first anomaly value type; Based on the historical feature data, obtain the second abnormal numerical type and the second occurrence frequency that appear in the sequential time periods, and determine the second correlation value of the second abnormal numerical type based on the second occurrence frequency of the second abnormal numerical type; Based on the first correlation value and the second correlation value, establish a type correlation rule, and perform type correlation on the abnormal numerical types based on the weighted correlation rule to obtain a type correlation result; Based on the abnormal level of the first abnormal numerical type, perform level weighting on the first correlation value to obtain a first level weighted value, and based on the abnormal level of the second abnormal numerical type, perform level weighting on the second correlation value to determine a second level weighted value; Based on the first level weighted value and the second level weighted value, combined with the abnormal numerical level, perform level weighting on the type correlation result to obtain a weighted correlation result; Based on the weighted correlation result, obtain an initial abnormal numerical weighted relationship distribution map; Based on the personalized information of the critically ill patients, determine the personalized weight for the abnormal type, and perform personalized weighting on the initial abnormal numerical weighted relationship distribution map based on the personalized weight to obtain an abnormal numerical weighted relationship distribution map.

8. The intensive care life information monitoring method applying a machine learning algorithm according to claim 6, characterized in that Generating a warning reminder for medical staff based on the abnormal numerical weighted relationship distribution map includes: Obtain the abnormal type level, weighted correlation value and distribution range in the abnormal numerical weighted relationship distribution map; Judge whether the abnormal type level is within the preset range; If so, based on the abnormal type, combined with the weighted correlation value and the distribution range, determine the potential health risk, set a low warning level for the potential health risk, and generate a warning reminder for medical staff; Otherwise, based on the abnormal type, combined with the weighted correlation value and the distribution range, determine the health abnormality information, set a high warning level for the potential health risk, and generate a warning reminder for medical staff.

9. The intensive care life information monitoring method applying a machine learning algorithm according to claim 1, characterized in that, It also includes real-time optimization of the patient physical sign abnormality recognition model, including: Obtain the abnormal prediction result of the patient physical sign abnormality recognition model within a time period, and obtain the actual diagnosis result, and obtain the result difference between the abnormal prediction result and the actual diagnosis result; Obtain the first physical sign data corresponding to all result differences, obtain the data characteristics of the first physical sign data, determine the proportion of the data characteristics in the historical feature data, and judge whether the proportion is greater than the preset proportion; If so, do not adjust the historical physical sign data; Otherwise, adjust the historical physical sign data based on the preset proportion to obtain updated historical physical sign data; Obtain the second physical sign data with a result difference greater than the preset result difference, obtain the historical marked feature of the second physical sign data in the historical feature data, correct the historical marked feature based on the result difference to obtain a target marked feature, and update the mark of the updated historical physical sign data based on the target marked feature to obtain new training data; Configure the first attention head of the attention mechanism for the adjusted physical sign data in the new training data, and configure the second attention head of the attention mechanism for the physical sign data with the target marked feature in the new training data; Based on the first attention head and the second attention head, an attention mechanism is added inside the model, and the abnormal sign recognition model is retrained and optimized with new training data to obtain a real-time abnormal recognition model.

10. The intensive care life information monitoring method applying a machine learning algorithm according to claim 9, characterized in that Based on a preset ratio, the historical sign data is adjusted to obtain updated historical sign data, including: Adjusting the data characteristics of the first sign data to have a ratio greater than the preset ratio in the historical sign data to obtain the removed data volume; Obtaining the data content in which no differences in all results appear in the historical sign data, and uniformly removing the data content according to the removed data volume to obtain updated historical sign data.

Images